1. Field of the Invention
The present invention relates to a method and an apparatus for detecting inadequately trimmed images from a large amount of trimmed images obtained through automatic trimming. The present invention also relates to a program that causes a computer to execute the method for detecting inadequately trimmed images.
2. Description of the Related Art
In the case of application for a passport or driver's license or generation of a resume, submission of a photograph representing a face (hereinafter referred to as an ID photo) is often required in a predetermined format. For this reason, an automatic photography apparatus has been used. Such an automatic photography apparatus has a photography room and photographs a user sitting on a stool in the room. Face images of the user are then printed on a sheet. In this manner, the sheet of photo IDs is generated automatically. However, such an apparatus is large and locations for installation thereof are thus limited. Consequently, a user needs to find and visit the location to obtain an ID photo, which is inconvenient for the user.
In order to solve this problem, a method of generating an ID photo has been proposed, as in Japanese Unexamined Patent Publication No. 11(1999)-341272, for example. In this method, when a user specifies positions of the top of the head and the chin in a face image displayed on a display device such as a monitor, a computer finds a magnification ratio and a face position according to the two positions and a format of an ID photo to be generated. The computer then enlarges or reduces the face image so that the face in the image after enlargement or reduction can be positioned in a predetermined position in the ID photo to be generated. Thereafter, the computer trims the enlarged or reduced image, and generates the ID photo. By using this method, a user can request generation of an ID photo from a DPE store or the like, which is much easier to find than an automatic photography apparatus. In addition, a user can also request generation of an ID photo by bringing a film or a recording medium having a preferred photo recorded therein to a DPE store or the like.
However, in this method, an operator needs to specify the two positions in a displayed face image, which is a burden on the operator especially in the case where the number of users is large. Furthermore, in the case where an area of the face in a face image being displayed is small or in the case where resolution of a face image is low, the operator finds it difficult to promptly and accurately specify the two positions. Therefore, prompt generation of an appropriate ID photo cannot be expected.
For this reason, methods for promptly and accurately setting a trimming range while reducing the burden on an operator have also been proposed. Especially, following the progress of automatic detection techniques, an automatic trimming method has been drawing attention. In this method, a face and eyes are detected in a photo and a trimming range is set according to a result of detection. By using such a method, an ID photo can be generated from a face image without specification of the two positions by an operator. For example, a method has been proposed in U.S. patent Application Publication No. 20020085771 for setting a trimming range. In this method, positions of the top of the head and eyes are detected in a face image and a position of chin is inferred based on the detected positions. In this manner, the trimming range is set.
Meanwhile, in the case of updating IDs of employees in a large company or in the case of issuance or update of driver's licenses in a driver's license center, efficient processing thereof is desired. For example, an operation is not efficient if each applicant is photographed to obtain a face image and a trimmed image is generated from the face image for generation of an ID or a driver's license (hereinafter collectively referred to as a card). However, if an operation is divided into photography processing, trimming processing, and card generation processing, all applicants can be photographed to obtain face images, and trimmed images can be generated from the face images and used for card generation. In this manner, operators dedicated to each processing can be trained, which is efficient.
However, although automatic trimming has a high rate of success, detection of faces and eyes in photographs is not necessarily 100% successful. For example, in the case where a person in a photograph to be subjected to automatic trimming has a unique hairstyle or wears tinted glasses, failure of detection may lead to wrong automatic trimming range setting or a wrong area size regarding a part to be trimmed automatically (the face in the case of a human). Therefore, inspection by an operator is desired regarding a large amount of trimmed images generated through automatic trimming.
An operator can inspect trimmed images by viewing the trimmed images displayed one by one, for example. This method is simple and thus easy to carry out. However, the time necessary therefore is as long as in the case of trimming carried out by an operator, and a burden on the operator is heavy. Therefore, in order to shorten the time for inspection, a method may be carried out for detecting inadequately trimmed images by display of thumbnail images generated by reduction of trimmed images. In this method, the thumbnail images are inspected for presence of a different size or position of a part to be trimmed. However, although this method can shorten the time by inspecting the images collectively, using the thumbnail images generated from the trimmed images lowers accuracy of detection.